Elsevier

NeuroImage

Volume 59, Issue 2, 16 January 2012, Pages 1668-1683
NeuroImage

Imitation and observational learning of hand actions: Prefrontal involvement and connectivity

https://doi.org/10.1016/j.neuroimage.2011.09.021Get rights and content

Abstract

The first aim of this event-related fMRI study was to identify the neural circuits involved in imitation learning. We used a rapid imitation task where participants directly imitated pictures of guitar chords. The results provide clear evidence for the involvement of dorsolateral prefrontal cortex, as well as the fronto-parietal mirror circuit (FPMC) during action imitation when the requirements for working memory are low. Connectivity analyses further indicated a robust connectivity between left prefrontal cortex and the components of the FPMC bilaterally. We conclude that a mechanism of automatic perception–action matching alone is insufficient to account for imitation learning. Rather, the motor representation of an observed, complex action, as provided by the FPMC, only serves as the ‘raw material’ for higher-order supervisory and monitoring operations associated with the prefrontal cortex. The second aim of this study was to assess whether these neural circuits are also recruited during observational practice (OP, without motor execution), or only during physical practice (PP). Whereas prefrontal cortex was not consistently activated in action observation across all participants, prefrontal activation intensities did predict the behavioural practice effects, thus indicating a crucial role of prefrontal cortex also in OP. In addition, whilst OP and PP produced similar activation intensities in the FPMC when assessed during action observation, during imitative execution, the practice-related activation decreases were significantly more pronounced for PP than for OP. This dissociation indicates a lack of execution-related resources in observationally practised actions. More specifically, we found neural efficiency effects in the right motor cingulate–basal ganglia circuit and the FPMC that were only observed after PP but not after OP. Finally, we confirmed that practice generally induced activation decreases in the FPMC during both action observation and imitation sessions and outline a framework explaining the discrepant findings in the literature.

Highlights

► Prefrontal cortex was involved in a rapid imitation task with low memory load. ► Connectivity confirmed between prefrontal cortex and mirror neuron system. ► Prefrontal cortex activations were correlated with observational practice effects. ► Observational practice lacked neural efficiency effects in putamen and cingulate. ► We confirmed practice-induced activation decreases in imitation and observation.

Introduction

Imitation learning refers to the learning of motor actions guided by observing a skilled model. Typically, imitation learning consists of alternating phases of observation and of motor execution, whereas the term observational practice (Vogt, 1995) refers to learning by observing in the absence of motor execution (e.g., Mattar and Gribble, 2005). Over the last 15 years, a large number of neuroimaging studies have assessed the brain networks underlying the imitation of relatively simple, familiar actions, whereas to date, only a handful of studies have explored the more complex imitation learning.

With respect to familiar actions, the meta-analysis by Caspers et al. (2010) confirmed that both action observation and imitative execution involve a bilateral network within ventral premotor and inferior parietal cortex. This fronto-parietal ‘mirror’ circuit (FPMC, Rizzolatti and Sinigaglia, 2010) has become the subject of intense debate, particularly regarding its role in action recognition and conceptual processing (Gallese et al., 2011, Hickok and Hauser, 2010, Kalenine et al., 2010, Kilner, 2011, Mahon and Caramazza, 2008). In contrast, the view that the engagement of the FPMC during action observation is functionally related to the engagement of the same regions during subsequent imitative execution (e.g., Vogt, 2002, Vogt et al., 2007) has not been challenged to date. Indeed, experiments using transcranial magnetic stimulation (TMS) over regions of the FPMC indicate their crucial role in imitation performance (e.g., Heiser et al., 2003).

With respect to actions that are not in the observer's behavioural repertoire, recent neuroimaging studies from our group revealed three key findings: first, the FPMC was activated more strongly during observation of non-practised actions compared to practised actions (Vogt et al., 2007; see Discussion for other, potentially conflicting findings). Second, the activations in the FPMC were stronger during action observation in order to imitate than during passive action observation (Buccino et al., 2004; see also Frey and Gerry, 2006). Third, in addition to the FPMC, the dorsolateral prefrontal cortex (DLPFC) was also activated during action observation and motor preparation (Buccino et al., 2004, Vogt et al., 2007). As in the FPMC, activations in DLPFC were also more pronounced for novel than for practised actions. The involvement of DLPFC has inspired a hierarchical model of imitation learning (ibid., see also Iacoboni, 2009, Ferrari et al., 2009), according to which left DLPFC engages in supervisory operations of selection and combination of the elementary action representations in the FPMC, and right DLPFC engages in monitoring operations during imitative execution.

The first aim of the present study was to further elucidate the role of the DLPFC in imitation learning. In their dual-route model of imitation, Ferrari et al. (2009) proposed that the ‘indirect mirror pathway’ (involving FPMC and DLPFC) is crucial (1) in tasks with delayed behavioural responses that require the maintenance of motor information beyond the available visual input, as well as (2) for behavioural parsing and recombination of motor elements, and (3) inhibitory control. These suggestions were partly based on our previous studies, in which observation and execution were separated by a relatively long preparatory pause (4 to 10 s). It is thus possible that the activations in DLPFC reflected predominantly the requirement to maintain motor representations in the absence of visual input, that is, explanation (1) above. Also Johnson-Frey et al. (2005) interpreted their activation in left DLPFC during the (non-imitative) planning of tool-use actions in this way. In addition, specific evidence for the role of DLPFC for supervisory and monitoring functions, rather than working memory per se, is only available from non-imitative tasks (e.g., Cunnington et al., 2006, Lau et al., 2004, Rowe and Passingham, 2001, Rowe et al., 2010). Therefore, we aimed to assess if prefrontal activations are present in the initial stage of imitation learning specifically when the task demands for maintenance of motor information are minimised. To this end, we used a rapid imitation paradigm in the present study: in each trial, participants were shown a picture of the to-be-imitated hand posture (guitar chord), which they executed immediately. In addition to this IMI condition, we included separate OBS sessions with the same trial structure, except that participants did not execute the chord during observation. Furthermore, the scanning parameters of the present study were optimised to allow for functional connectivity analyses between the DLPFC and the two key regions of the FPMC. This connectivity is a crucial, but yet untested assumption of the hierarchical model of imitation learning (Buccino et al., 2004).

The second aim of the present study was to contrast the neural correlates of two forms of practice: imitation learning, where action observation is immediately followed by imitative execution (‘Physical Practice', PP), and learning by observing without motor execution (‘Observational Practice', OP; for reviews of behavioural studies, see Maslovat et al., 2010, Vogt and Thomaschke, 2007). Regarding the possible neural network underlying OP of complex, novel actions, two accounts are feasible that differ regarding the involvement of the DLPFC. The first account is a straightforward extension of Buccino et al.'s (2004) hierarchical model of imitation learning to observational practice. This assumes the involvement of the indirect pathway (FPMC under the control of DLPFC) also for periods of OP. However, a number of findings tentatively contradict this account. In our previous studies, DLPFC was only found to be activated when action observation was followed by imitative execution within the same trial (Buccino et al., 2004: IMI condition; Vogt et al., 2007). In contrast, no DLPFC activations were found during trials (Buccino et al., 2004: OBS condition) or sessions (Frey and Gerry, 2006) where action observation was not followed by execution, even when participants watched novel, complex actions with the intention to imitate these after scanning (Cross et al., 2009, Cunnington et al., 2006, Frey and Gerry, 2006). The only available evidence in support of the first account comes from the study by Torriero et al. (2007), who demonstrated that repetitive TMS over the right DLPFC prior to observational practice of a spatial sequencing task selectively disrupted subsequent reproduction of the practised sequence.

The second account assumes that Ferrari et al.'s (2009) direct pathway, largely consisting of the FPMC, is sufficient to explain the effects of OP of simple, as well as of more complex, novel actions. For simple actions, Stefan et al., 2005, Stefan et al., 2008; see also Celnik et al., 2006) demonstrated longer-term effects of observational practice on TMS-evoked thumb movements, which are most likely underpinned by downstream activations from the FPMC to primary motor cortex (M1) during action observation (Kilner et al., 2009, Tkach et al., 2007). A similar conclusion was reached by Brown et al. (2009), who demonstrated that repetitive TMS over M1 applied after observational practice reduced the practice effects in learning a novel force environment (see also Malfait et al., 2010). In the present study, the OBS scanning sessions were used to assess the functional activations during OP. We further ran correlation analyses between the functional data and the behavioural learning outcomes for a full assessment of the involvement of the DLPFC in action observation.

Our second motivation for studying physical and observational practice conditions (Aim 2.2) was to identify specific patterns of neural activation, in OBS and IMI sessions, that might result from each form of practice. To our knowledge, only two fMRI studies have pursued this approach to date: Nyberg et al. (2006) contrasted the effects of PP and motor imagery of finger-tapping sequences. Relative to non-practised sequences, participants who had engaged in PP exhibited stronger activations in the supplementary motor area (SMA) and the cerebellum, whereas participants who had engaged in motor imagery exhibited stronger activations in visual association cortex (BA 18). The authors interpreted their findings to indicate training-specific neuroplastic changes. In the second study (Cross et al., 2009), participants passively observed one set of dance sequences and engaged in PP with another set. Overall, the results indicated a strong overlap between the danced and watched conditions in the regions of the FPMC, and the small differences obtained can be attributed to the absence of an instruction to learn the observed sequences. In contrast, participants in the present study knew that they would be tested for both the physically and the observationally practised actions. They were further encouraged to engage in motor imagery simultaneously and/or following action observation in order to maximise the possible beneficial effects of learning by observing. In addition, our study is the first to examine functional activations not only during observation of the differently practised actions (OBS sessions), but also during imitative execution (IMI sessions).

The third aim of the present study was to investigate practice effects in the guitar chord paradigm over a slightly longer period than previously studied. Whereas a number of earlier studies on expertise effects on action observation found stronger activations for motor experts or for practised actions (see Discussion), Vogt et al. (2007) found reduced activations for observation of practised relative to novel actions. To assess the reliability and stability of this finding, participants in the present study were scanned after one day of practice (on Day 2, as in Vogt et al., 2007) and they were scanned again (on Day 4) after a further day of practice. The scanning data from Day 4 also served as the main database for the comparisons between the OP- and PP-actions (Aim 2.2).

To summarise, in the present study participants practised one set of guitar chords via observational practice (OP) and another set via imitative execution (PP) in a within-subjects design over two practice sessions (Days 1 and 3). The scanning sessions on Days 2 and 4 comprised IMI blocks where participants imitated the PP-, OP-, as well as non-practised chords, and OBS blocks where participants observed these chord types. The study was designed with the following aims in mind:

  • (1)

    to assess the involvement of prefrontal cortex in the initial stage of imitation learning, as well as its connectivity to the FPMC, using a rapid imitation paradigm that minimised requirements for maintenance of motor information;

  • (2.1)

    to explore the neural correlates of observational practice and specifically the involvement of the DLPFC during action observation;

  • (2.2)

    to identify patterns of activation, during observation and imitation sessions, that might distinguish actions that had been practised by OP vs. by PP; and

  • (3)

    to assess the reliability of activation differences between practised and non-practised actions as reported in Vogt et al. (2007), as well as to assess their stability after further practice.

Section snippets

Participants

Eighteen healthy, right-handed volunteers (aged 24.2 ± 3.8 years) participated in the study. All participants had normal or corrected-to-normal visual acuity, and no participant had experience with string musical instruments. One participant was excluded from the data analysis due to lack of practice effects in the behavioural data. Participants gave their written informed consent to the experimental procedure, which was approved by the local Ethics committee.

Design and experimental conditions

All participants attended two practice

Behavioural data

Fig. 2 shows the participants' response times, measured from chord display onset to stable finger positioning on the fretboard and averaged across the two IMI blocks on each scanning day. Prior to analysis, trials with late or no hand movements were excluded. These amounted to only 2.6% each for the NP- and OP-trials, and only 1 trial was excluded for the PP- and S-trials. We confined the analysis to the chronometric data due to the ultimately subjective nature of error scoring in this task. A

Discussion

The results of this study inform three areas of research, one concerned with the role of prefrontal cortex in imitation learning, the second concerned with the neural correlates of learning by observing, and the third with neural efficiency and practice more generally. In short, we have presented clear evidence for the involvement of the DLPFC in a complex imitation task with low demands on working memory (Aim 1, see Introduction), and we have demonstrated a robust functional connectivity

Conclusions

Heyes (2009) noted “a growing body of theory and evidence suggesting that mimicry and imitation learning are continuous”, and that “imitation learning enlists additional, general purpose mechanisms of learning and cognitive control” (p. 2295). The present results support both proposals, in that we indeed found the FPMC involved in imitation learning, and that prefrontal cortex is most likely exerting a supervisory and monitoring role over the elementary representations provided by the FPMC,

Acknowledgments

We would like to thank Valerie Adams, Anna Anderson, and Bill Bimson (Liverpool), as well as Dave Gaskell and Gordon Johnson (Lancaster) for their kind technical and administrative support during setup and running of this study. Virginia Kellond (Liverpool) helped with participant-management and with scoring the video-recordings, and Daniel Eaves (Teesside and Lancaster) and Peter Walker (Lancaster) helped improve the English. Special thanks go to Brent Vogt (New York) for his invaluable advice

References (93)

  • R. Cunnington et al.

    The selection of intended actions and the observation of others' actions: a time-resolved fMRI study

    Neuroimage

    (2006)
  • S.B. Eickhoff et al.

    A new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data

    Neuroimage

    (2005)
  • S.B. Eickhoff et al.

    Testing anatomically specified hypotheses in functional imaging using cytoarchitectonic maps

    Neuroimage

    (2006)
  • S.B. Eickhoff et al.

    Assignment of functional activations to probabilistic cytoarchitectonic areas revisited

    Neuroimage

    (2007)
  • K. Emmorey et al.

    CNS activation and regional connectivity during pantomime observation: no engagement of the mirror neuron system for deaf signers

    Neuroimage

    (2010)
  • F. Filimon et al.

    Human cortical representations for reaching: mirror neurons for execution, observation, and imagery

    Neuroimage

    (2007)
  • K.J. Friston et al.

    Psychophysiological and modulatory interactions in neuroimaging

    Neuroimage

    (1997)
  • K.J. Friston et al.

    How many subjects constitute a study?

    Neuroimage

    (1999)
  • K.J. Friston et al.

    Dynamic causal modelling

    Neuroimage

    (2003)
  • D.R. Gitelman et al.

    Modeling regional and psychophysiologic interactions in fMRI: the importance of hemodynamic deconvolution

    Neuroimage

    (2003)
  • U. Halsband et al.

    Motor learning in man: a review of functional and clinical studies

    J. Physiol. Paris

    (2006)
  • G. Hickok et al.

    (Mis)understanding mirror neurons

    Curr. Biol.

    (2010)
  • O. Hikosaka et al.

    Parallel neural networks for learning sequential procedures

    Trends Neurosci.

    (1999)
  • O. Hikosaka et al.

    Central mechanisms of motor skill learning

    Curr. Opin. Neurobiol.

    (2002)
  • M. Iacoboni

    Neurobiology of imitation

    Curr. Opin. Neurobiol.

    (2009)
  • H. Imamizu et al.

    Reorganization of brain activity for multiple internal models after short but intensive training

    Cortex

    (2007)
  • J. Jankowski et al.

    Distinct striatal regions for planning and executing novel and automated movement sequences

    Neuroimage

    (2009)
  • J.M. Kilner

    More than one pathway to action understanding

    Trends Cogn. Sci.

    (2011)
  • A.J. Kovacs et al.

    Coding of on-line and pre-planned movement sequences

    Acta Psychol.

    (2010)
  • H.C. Lau et al.

    Willed action and attention to the selection of action

    Neuroimage

    (2004)
  • D.R. Leff et al.

    Changes in prefrontal cortical behaviour depend upon familiarity on a bimanual co-ordination task: an fNIRS study

    Neuroimage

    (2008)
  • B.Z. Mahon et al.

    A critical look at the embodied cognition hypothesis and a new proposal for grounding conceptual content

    J. Physiol. Paris

    (2008)
  • A.G. Mattar et al.

    Motor learning by observation

    Neuron

    (2005)
  • J. Milton et al.

    The mind of expert motor performance is cool and focused

    Neuroimage

    (2007)
  • L. Nyberg et al.

    Learning by doing versus learning by thinking: an fMRI study of motor and mental training

    Neuropsychologia

    (2006)
  • S. Panzer et al.

    Inter-manual transfer and practice: coding of simple motor sequences

    Acta Psychol.

    (2009)
  • A. Roebroeck et al.

    Mapping directed influence over the brain using Granger causality and fMRI

    Neuroimage

    (2005)
  • J.B. Rowe et al.

    Working memory for location and time: activity in prefrontal area 46 relates to selection rather than maintenance in memory

    Neuroimage

    (2001)
  • M.B. Schippers et al.

    Mapping the flow of information within the putative mirror neuron system during gesture observation

    Neuroimage

    (2011)
  • S. Vogt et al.

    Prefrontal involvement in imitation learning of hand actions: effects of practice and expertise

    Neuroimage

    (2007)
  • L.A. Bates et al.

    Imitation: what animal imitation tells us about animal cognition

    Wiley Interdiscip. Rev. Cogn. Sci.

    (2010)
  • N. Bien et al.

    The brain's intention to imitate: the neurobiology of intentional versus automatic imitation

    Cereb. Cortex

    (2009)
  • L.E. Brown et al.

    Repetitive transcranial magnetic stimulation to the primary motor cortex interferes with motor learning by observing

    J. Cogn. Neurosci.

    (2009)
  • R.W. Byrne

    Imitation as behaviour parsing

    Philos. Trans. R. Soc. Lond. B Biol. Sci.

    (2003)
  • R.W. Byrne

    Parsing behaviour: a mundane origin for an extraordinary ability?

  • B. Calvo-Merino et al.

    Action observation and acquired motor skills: an fMRI study with expert dancers

    Cereb. Cortex

    (2005)
  • Cited by (79)

    • Simultaneous and alternate combinations of action-observation and motor imagery involve a common lower-level sensorimotor process

      2022, Psychology of Sport and Exercise
      Citation Excerpt :

      Indeed, research from the motor learning literature consistently highlights a benefit of having the opportunity to deliberate or contemplate previous trial attempts (Guadagnoli and Kohl, 2001; Swinnen, Schmidt, Nicholson, & Shapiro, 1990), or switching between different routines or variations of a set task (Hall, Domingues, & Cavazos, 1994; Shea & Morgan, 1979). Moreover, the potential for motor learning through AO and/or MI is additionally accompanied by the recruitment of the lateral prefrontal neural regions, which are synonymous with attentional processes including action monitoring and movement reorganization (Buccino, Binkofski, & Riggio, 2004; Higuchi, Holle, Roberts, Eickhoff, & Vogt, 2012; see also, Eaves, Behmer, & Vogt, 2016). Thus, it is further possible to capture the learning or adaptation from AO and/or MI by also recognizing the importance of events that take place in between each of their presentations.

    View all citing articles on Scopus
    View full text